Thread: [Announce] modRana: a flexible GPS navigation system
View Single Post
misiak is offline misiak
2013-11-26 , 20:24
Posts: 805 | Thanked: 1,605 times | Joined on Feb 2010 @ Gdynia, Poland
#1510
Originally Posted by misiak View Post
I'll have to download another ~50km long route next week, I'll reswap (using your ereswap tool btw) before using modrana, then download tiles and use ereswap again (maybe modrana is using so much ram that it gets swapped to hell - then I should get the message that we are >9k over swap borderline and reswapping is needed), I'll post results then (although I doubt that 80mb of tiles could fill my 768 swap completely more than once, but you can never be sure unless you check).
Today I had to download 20 mb of tiles, and I did this result. The steps (and results) follow:
  1. launched ereswap, reswapping was performed
  2. launched ereswap once again, noted the value of what was left, let's call it "X" (it was almost 500 mb)
  3. launched modrana, waited for it to get gps fix and load tiles around my current position
  4. launched ereswap again - the "left" value shrinked by 60mb - which means starting modrana uses 60mb of swap
  5. entered route (address to address), searched it via google and started navigation, chosen option to download tiles around my current route - it was ~15k tiles, summing up to ~20mb, began downloading, stopped after 1/3
  6. launched ereswap - the "left" value shrinked by 30mb while downloading <7mb of tiles (but I guess it could be also caused by the routing start)
  7. chosen option to download tiels around my current route again - ~10k tiles left, suming up to ~13mb, paused at half (so that was again 1/3 of whole batch)
  8. launched ereswap - while downloading <7mb of tiles this time "left" value shrinked by almost 15mb
  9. downloaded the rest of tiles, launched ereswap and noted the value - 10-15 mb this time
  10. closed modrana

Please note that modRana and xterm were only applications running, no incoming calls, no sms messages, nothing more. The tiles were downloaded via wifi, kernel-power 52, no bleeding edge drivers. I guess the only thing using swap in large chunks was modRana then. Someone should repeat my tests to be sure, but my experience is that modRana swap usage during tile download is more or less 2x size of downloaded tiles (I'm ignoring the first 1/3 batch here, as the routing start could also take ram/swap and interfere with the measurments). In my opinion, that shouldn't make the phone freeze, so the bottleneck must be something else.

edit: I trust the ereswap used swap measurement method is correct, I haven't analysed the code (just pointing that out, as it could also have influence on the result and its credibility).

edit2: I used N900's stock swap configration for this test (e.g. /proc/sys/vm/swappiness set to "100").
Last edited by misiak; 2013-11-27 at 14:25. Reason: added information about using stock swap configuration
 

The Following 2 Users Say Thank You to misiak For This Useful Post: